Re: Truth Machines and Open Networks (and FI)

Craig Presson (dhr@iname.com)
Thu, 12 Feb 1998 08:16:25 -6


On 12 Feb 98 at 6:10, Brent Allsop wrote:

> Great Post Wax <yakwax@yahoo.com>!
>
> > So would you mind giving up privacy for absolute truth?
>
> YES!
>

I'd mind, too. I finally ordered (through the alcor.org link to
Amazon -- Amazon's one-click ordering is _dangerous_!) and read _The
First Immortal_ -- now I have to go to the archives and read the
threads I skipped because they had FI spoilers!

I agree FI is a terrific read. I sucked it down, using all my spare
time out of about 30 calendar hours earlier this week. I even missed
some sleep for it. Found a distinct Heinleinian flavor to it at the
end ...

IF a truth machine is possible -- and given some of the skilled liars I
have known, I wonder -- it could be used, per Halperin, to put an end
to at least part of politics-as-usual, and to provide assurance of good
faith in contracts (insert digression about how to implement truth
certificates on the Net, since many contracts are made without f2f
contact). However, in the context of the book it felt to me like a Deus
ex Machina -- enter the truth machine, exit the gray-goo and rogue-AI
problems. I'd still prefer a technical solution to the former, and will
have to remain skeptical about the latter. Halperin has it both ways in
FI: his AIs start with no survival imperative or emotions, yet they
evolve. To be sure those traits are never introduced, you'd have to
prove one of two propositions:

1. That those traits can't evolve from AIs that don't start with them;
2. That a machine can recognize those traits in another machine it is
building, so that it can choose another design approach to avoid
introducing them.

My gut feeling is that 1 is not true and 2 is difficult.

Further, I think emotion, or some analogue of it, is tied in to
intelligence at fundamental levels, and that Halperin's
perfectly-Apollonian AIs would have to be able to at least model human
emotion and drives to do what they do (create art and act as judges,
for two instances). (I used to have some references on this, but I
think my Labrador ate them. I apologize for glossing over such an
interesting point).

Well, the book provides a lot of food for thought. I did catch a couple
of minor technical flaws, but I will be surprised if I am the first to
see them.

I have to cut this short, being still a member of the involuntary work
force <grin>.

-- dhr@iname.com (Freeman Craig Presson)